Circular RNAs (circRNAs) are closed-loop single-stranded RNA molecules that have proved to be important regulators of gene expression at multiple levels although initially considered transcriptional byproducts [90]. CircRNAs function as miRNA sponges or traps that indirectly modulate transcription, interact with intracellular proteins, regulate splicing, and travel in extracellular vehicles called exosomes, enabling intercellular communication [91,92]. In the context of normal hematopoiesis, circRNAs show cell-type specificity and are considered as regulators of blood cell differentiation and maturation [93].

The hypothesis of circRNAs interfering with MDS pathophysiology was supported by the observation that exogenous inhibition of the spliceosome components, commonly affected by MDS mutated genes, can cause an imbalance between circular and linear RNA concentrations within affected cells towards overexpression of the circular molecules [94,95]. Wedge et al. recently reported that specific cancer-associated circRNAs, such as circZNF609 and circCSNK1G3, are upregulated in MDS patients with U2AF1 mutations compared to unmutated controls [96]. Additionally, global circRNA expression has been found to be upregulated in the continuum from normal hematopoiesis to clonal cytopenias of undetermined significance (CCUS) and further to MDS. Even among MDS patients, a higher risk group was correlated with increased global circRNA expression and a “Myeloid Circ Score” was developed based on 14 specific circRNAs with potential prognostic value to stratify patients in terms of risk and disease outcomes [97]. Another research group found 145 circRNAs to be upregulated and 224 downregulated in MDS patients compared to healthy controls. Researchers also suggested that of all these circRNAs, hsa_circRNA_100352, hsa_circRNA_104056, and hsa_circRNA_102817 could be used as MDS prognostic biomarkers, since their increased expression was significantly correlated with poorer OS. Bioinformatics network analysis indicated that these three circRNAs are probably associated with multiple cancer-related molecular pathways, including Wnt/β-catenin and PTEN/Akt/mTOR [98,99]. Additionally, circ-ANAPC7 might be another promising circRNA biomarker, as its expression in MDS patients has recently been shown to be upregulated, along with the increasing risk group, by IPSS-R [100]. Finally, several circRNAs are differentially expressed between responders and nonresponders to azacytidine, although only one circRNA, hsa_circ_0006595, is considered a potential predictor for response to azacytidine treatment [101]. Whether circRNAs will soon be used in clinical practice for diagnostic, prognostic, or predictive purposes remains to be answered, given the need for bone marrow sampling, since the reproducibility of findings in peripheral blood has not been proven yet.

Long non-coding RNAs (lncRNAs) are a functionally heterogeneous class of thousands of RNA molecules, each containing more than 200 nucleotides, which are not translated into functional proteins. They are produced through DNA transcription, either from genes or intergenic regions (lincRNAs), and have multiple functions including epigenetic chromatin modifications, regulation of neighboring and distant gene transcription, RNA splicing, response to DNA damage, sponging miRNAs, and participation in signaling pathways [102,103]. In the field of normal hematopoiesis, from murine models to humans, it is known that lncRNAs are expressed in a stage-specific and lineage-specific pattern from hematopoietic stem cells (HSCs) to mature blood cells in a way that they enable self-renewal of HSCs, such as H19 lncRNA, but also determine lineage commitment of progenitor cells, e.g., EGOT lncRNA for eosinophil maturation, in cooperation with transcription factors [104,105,106,107,108,109,110].

After the identification of MEG3 (maternally expressed gene 3) lncRNA hypermethylation in many MDS patients, evidence that linked aberrant expression of lncRNAs with multiple hematological malignancies, including MDS, began to accumulate. The aforementioned lncRNA is considered a tumor suppressor whose downregulation has been associated with poor OS in several solid neoplasms [111,112,113,114,115,116]. While scientific interest in lncRNAs was increasing, researchers identified a positive feedback loop in MDS cells involving lncRNA bc200-miR-150-5P-MYB, which resulted in sustained cell proliferation. On the other hand, the inhibition of this axis seemed to suppress neoplastic growth of bone marrow MDS cells, implying potential therapeutic targeting of BC200 [36]. Additionally, increased expression of the lncRNAs KCNQ10T1 and HOXB-AS3 has been associated with adverse prognosis in MDS, with the latter pertaining to only lower-risk patients [117,118]. Further basic research and computational analysis revealed a vast number of differentially expressed lncRNAs between MDS patients and healthy controls, with functions including cell adhesion, differentiation, and chromatin modifications, mainly through functional interaction with DNA methylation processes [119,120]. Of these lncRNAs, H19 emerged as one of the most promising prognostic biomarkers in MDS patients. Interestingly, a set of 14 lncRNAs were considered as reliable predictive biomarkers to inform about potential patients’ response to azacytidine [101,120,121]. To improve MDS risk stratification by connecting laboratory research with clinical practice, Yao et al. developed a scoring system based on the expression of four lncRNAs with the highest prognostic potential (TC07000551.hg.1, TC08000489.hg.1, TC02004770.hg.1, TC03000701.hg.1). A higher lncRNA score was significantly associated with higher bone marrow blast percentage, higher-risk subtypes by WHO, complex karyotypes, high-risk gene mutations (RUNX1, ASXL1, TP53, SRSF2, and ZRSR2), as well as shorter OS [122]. Consequently, lncRNAs overall appear to be promising prognostic and predictive biomarkers for patients with MDS, probably awaiting their future incorporation in widely accepted prognostic scoring systems to assist in decision-making.

PIWI-interacting RNAs (piRNAs), the third major class of small non-coding RNAs, are single-strand 26–31 nucleotide-long RNA molecules. Their main function, apart from epigenetic modifications, was first believed to be the maintenance of germline DNA integrity through the guidance of PIWI proteins (P-element-induced wimpy testis proteins) towards silencing transposons, which are mobile parasitic genomic elements [123,124]. Further research indicated that aberrant expression of specific piRNAs is associated with the development and progression of several solid and hematological cancers, as these molecules are considered to play a role in continuous proliferative signaling, resistance to apoptosis, tumor invasion, angiogenesis of malignant tissues, and even resistance to antineoplastic treatment [125,126]. On the other hand, though, there has been increasing evidence that aberrant expression of piRNA pathway genes alone might not be adequate for the formation of piRNA–PIWI silencing complexes with biological impact on tumorigenesis [127].

Although the importance of piRNAs in other hematological malignancies such as multiple myeloma and classic Hodgkin lymphoma has gathered research interest, data on MDS have been scarce. The first study of piRNAs in bone marrow cells of patients with MDS demonstrated a higher expression (9%) of piRNAs in patients with MDS with refractory anemia (low-risk MDS) compared to patients with MDS with refractory anemia and excess of blasts—2 (high-risk MDS) and healthy controls (2% and 1%, respectively), assuming a DNA-protective role of piRNAs in lower-risk MDS [128,129]. Small non-coding RNA analysis from plasma and extracellular vesicles also showed an upregulation of specific piRNAs (hsa_piR_019914/gb/DQ597347 and hsa_piR_020450/gb/DQ598104) in MDS patients compared to controls. Two other piRNAs, hsa_piR_000805/gb/DQ571003 and hsa_piR_019420/gb/DQ596670, were differentially expressed between patients with low- and increased blasts—MDS. The latter piRNA was also shown to be correlated with OS with a protective role, but no piRNAs were found to have predictive value about patients’ response to azacytidine [76]. The biologic interpretation of these findings as well as the extent to which they can be incorporated in everyday clinical practice remain to be further elucidated.

Ribosomal RNAs (rRNAs) are indispensable components of ribosomes, the cell’s protein-producing machinery. Ribosomes in human cells comprise four rRNAs (28S, 5S, 5.8S, and 18S) and approximately 80 proteins that are assembled into a small (40S) and a large (60S) subunit through a multilevel process, which mainly takes place in the nucleolus [130,131,132,133].

The dependence of highly proliferative cells, such as the hematopoietic cells, upon protein synthesis has provided the rationale for extensive research on the role of aberrant ribosomal synthesis in several human diseases including hematopoietic neoplasms. In this context, mutation of Nol9 a ribosomal biogenesis protein required for 28S rRNA processing was found to affect hematopoiesis in animal models by reducing proliferation of hematopoietic stem and progenitor cells [134]. Moreover, DNAJC21 mutations were associated with bone marrow failure with increased tendency to malignancy, attributed to impaired biosynthesis and cytoplasmic maturation of the 60S ribosomal subunit [135]. Similarly, a whole group of diseases termed “ribosomopathies” arising from congenital or acquired genetic abnormalities that lead to impaired ribosomal construction and function have been associated with bone marrow failure and increased risk of hematological malignancies, such as Shwachman–Diamond syndrome or congenital dyskeratosis [136,137]. Further data supporting the correlation of rRNA deregulation with myeloid neoplasms indicate the potential role of DDX41, whose germline mutations predispose to myeloid malignancies, in the processing of pre-ribosomal rRNA to mature rRNA [138]. U2AF1 somatic mutations, commonly detected in MDS patients, apart from altered splicing, are also believed to cause aberrant ribosomal synthesis, mediated by NPM1, which is considered a ribosomal biogenesis factor [139]. Finally, bone marrow CD34+ cells from patients with MDS show decreased rRNA expression compared to controls, which is probably driven by increased promoters’ methylation of DNA loci coding for these rRNAs (rDNA). Interestingly, this hypermethylation can be reversed by hypomethylating agents such as azacytidine and it is therefore implied that methylation status of rDNA could be used as a predictor of response to treatment with such agents, instead of genome-wide methylation status, although this hypothesis is yet to be proven [140,141]. Researchers have recently focused on the study of short RNA fragments cleaved from rRNA, called rRNA-derived fragments (rRFs), as they are believed to regulate cellular functions and show sequence overlap with miRNAs and piRNAs [142,143].

Small nucleolar (snoRNAs) RNAs are 60–300 nucleotide-long RNA molecules derived from coding and non-coding genes and are in the nucleolus of eukaryotic cells. Their main function is processing of other RNA molecules such as ribosomal RNAs and small nuclear RNAs (snRNAs) via pseudouridylation and 2′-O-methylation. In turn, snRNAs are vital components of the spliceosome, the cell machinery that catalyzes pre-mRNA splicing through intron excision and joining of exons, to form functional mature mRNAs [144]. Additionally, snoRNAs are involved in regulation of alternate splicing and also act like miRNAs to selectively suppress gene expression [145,146].

In HSCs, snoRNAs expression is supposed to be cell-type-specific and play an important role in cell homeostasis, self-renewal, and stress response, while their aberrant expression has been linked to several hematological malignancies, MDS included [147,148,149]. For example, DDX41 regulates snoRNA processing, ribosomal biogenesis, and protein synthesis in hematopoietic stem and progenitor cells (HSPCs) and its germline mutation is known to confer predisposition to clonal myeloid disorders. More specifically, monoallelic DDX41 mutations, as in germline predisposition, increase the risk for age-dependent hematopoietic defects and confer competitive proliferation advantage to HSPCs. On the other hand, biallelic DDX41 mutations deregulate snoRNA processing, causing intracellular accumulation of inappropriately processed snoRNAs; impair protein synthesis; and finally result in cell cycle arrest. Most of the affected snoRNAs belong to the SNORA family and are typically involved in RNA pseudouridylation [150]. Similarly, snoRNA U33, which is a mediator of cell metabolic stress, has been found to be upregulated in MDS patients. More importantly, this snoRNA was shown to be significantly associated with OS of patients, albeit no relevant biological explanation is provided [76,151].

Transfer RNAs (tRNAs), with their unique stem–loop pattern formed by internal base pairing, are essentially the carriers of amino acids to the growing polypeptide chain at the ribosome during translation but are also believed to have additional functions such as modulation of gene expression and control of cell death. Cleavage of pre- or mature tRNAs produces the tRNA-derived fragments (tRFs) or tRNA-derived small RNAs (tsRNAs) or tRNA-derived RNAs (tDRs), which are involved in multiple biological processes including translational regulation with gene silencing, intercellular communication, cellular stress response, and immune cell activation, rather than being useless byproducts of tRNA degradation [153,154,155,156].

Specific tRNAs (chr2.tRNA27-GlyCCC, chr.18Trna4-LysCTT) as well as overall tRNA to rRNA ratio have been found upregulated in marrow cells from MDS patients compared to controls, and it was assumed that this increase might contribute to decreased programmed cell death and increased leukemic transformation, since tRNAs are known to inhibit cytochrome c activated apoptosis [76,107,128,157]. On the other hand, the SF3B1^K700E mutation commonly seen in MDS seems to reduce translational machinery components, primarily tRNA synthetases [158]. Another somatic mutation in the mitochondrial tRNA repertoire, MtRNA^Leu(UUR), in bone marrow cells is suspected to contribute to ineffective hematopoiesis [159].

When it comes to tRFs, some of them show enhanced expression while others are downregulated in MDS cells. Interestingly, the combined expression of 4 tRFs (chr6.tRNA157.ValCAC, chr11.tRNA17.ValTAC, chrM.tRNA12.TS1, and chrX.tRNA4.ValTAC) in treatment-naïve patients was found to have predictive value regarding the likelihood of response to treatment, and this is also the case with one mitochondrial tRNA (MT-TSI), while it is suggested that tDR-Asp family members could be used as predictors for progression to AML [152,160].

Even posttranscriptional modifications of these non-coding RNAs are suspected to interfere with MDS pathophysiology. Pseudouridylation by PUS enzymes, for instance, of mini tRFs containing 5-terminal oligoguanine, was found to regulate the renewal of human embryonic stem cells and also promote the differentiation of impaired HSPCs in MDS, indicating a potential therapeutic approach [161,162,163].

Short or small interfering RNAs (siRNAs) are 21–25 nucleotide-long RNA molecules with a crucial role in gene silencing, primarily through mRNA degradation and by promoting heterochromatin formation. These interfering RNAs are produced via the procession of long double-stranded RNAs or short hairpin RNAs by the DICER endoribonuclease. The produced double-stranded siRNA is then packed with proteins to form the RISC. One strand of the RNA is discarded, and the remaining strand guides the RISC towards the targeted mRNA, which is recognized with perfect complementarity with the siRNA and is finally cleaved by Ago2 protein of the RISC [164,165,166].

The well-established way of action of RNA interference has not only made it possible for researchers to better understand its implications in cancer pathogenesis but also provided the possibility to utilize siRNAs towards gene expression knockdown with research and therapeutic purposes. For instance, siRNAs have been used in basic research as tools to knockdown expression of genes that are commonly mutated in MDS patients, such as ZRSR2 and antiapoptotic “survivin”, so as to better investigate their role in MDS pathophysiology [167,168]. Additionally, Mackin et al. showed that compared with azacytidine, which is a hypomethylating pharmacologic agent, siRNAs targeting DNMT expression (DNA methyltransferase) proved more efficient at overall demethylation within the genomic transcription units [169]. Another clue to the potential therapeutic role of siRNAs came when the siRNA-mediated inhibition of p38a MAP kinase, a mediator of apoptosis that is constitutively activated in low-risk MDS bone marrow cells, led to in vitro improvement of hematopoiesis from MDS myeloid and erythroid progenitors [170]. It is therefore implied that siRNAs could provide a means of therapeutically targeting multiple genes that are aberrantly expressed in MDS patients, although no such agents have been tested in MDS patients to date.